The televised address came on a Monday, and the president of South Korea, Lee Jae-myung, used it to declare 2026 the year the country would establish itself as an irreplaceable industrial power. The announcement that followed gave the declaration a number: more than $550 billion in commitments from the country’s largest technology companies to expand memory-chip production, a national response to a global shortage that has made DRAM the most strategically important commodity in the AI economy.
The commitments are led by Samsung Electronics and SK Hynix, the two companies that dominate the world’s memory markets alongside Micron Technology of the United States. The bulk of the money, about $518 billion, is earmarked for four new memory fabrication plants in the country’s southwest, with a further $52 billion for a high-bandwidth memory packaging hub in the central region. The program covers HBM4 development, advanced packaging and the AI data centers that will consume the output, and it follows a government plan unveiled the previous day.
The scale is difficult to exaggerate. Samsung separately said it will invest about 2,655 trillion won over the next decade, with a large share directed to the Honam region in the southwest, where it plans a new semiconductor fab in Gwangju and an AI data center in Haenam. The commitments bring South Korea’s total announced AI-related spending, including data centers to be built by conglomerates through 2035, to more than $900 billion. For a country of 51 million people, the sums are national-defense scale.
The reason for the urgency is visible in every data center on earth. High-bandwidth memory, the vertically stacked chips that feed AI accelerators, has become the bottleneck of the AI build-out, and the companies that make it hold pricing power that would have seemed absurd three years ago. Samsung and SK Hynix, along with Micron, effectively control the market, and the shortage has sent memory prices up sharply, inflating the costs of every computer, phone and server built this year.
The investment program is an attempt to convert that advantage into something permanent. The government and the companies say the new fabs and packaging lines will expand capacity, ease the shortage and secure South Korea’s position as the pricing authority for the AI era’s most constrained input. Officials have described the plan as ensuring that the country’s memory industry, which already generates a large share of its exports, remains indispensable as AI spreads from data centers to devices.
The competitive stakes are sharp. SK Hynix has led the industry in high-bandwidth memory, winning the earliest and largest orders from Nvidia, while Samsung has worked to close the gap after struggling with yields on early HBM generations. The new investment is, in part, Samsung’s answer: fresh capacity, a dedicated packaging hub and a long runway to match its rival’s position. Micron, meanwhile, has signed contracts for more than $100 billion of memory, and the United States has pushed to rebuild its own chip supply chain.
Executing at this scale is another matter. Four new fabs and a packaging hub require years of construction, thousands of engineers and enormous amounts of power and water, and the timeline for meaningful new capacity extends into 2027 at the earliest. The investments also carry the classic risk of the memory business: it is cyclical, and capacity built during a shortage has a history of arriving after the shortage ends, flooding the market just as demand softens.
The government is trying to manage that risk with policy. The president’s address promised incentives around power, water, workforce and living conditions for the southwest region, an attempt to spread the AI economy beyond the Seoul area and to make the new fabs economically viable. The program ties corporate investment to government support, a partnership that has characterized South Korea’s semiconductor policy for decades.
The program is also a statement about pricing power. Korea’s memory makers have used the shortage to renegotiate the terms of the memory business, with long-term contracts, prepayments and take-or-pay floors replacing the spot-market volatility of earlier cycles. The new investment is designed to lock in that structure: more capacity under contract, more pricing discipline and less exposure to the boom-and-bust that once defined the industry.
The investment lands in a geopolitical context that raises the stakes. The United States has pushed to rebuild its own memory industry, and China has poured resources into domestic chip production, and the competition for memory supremacy now runs through trade policy as much as technology. Korea’s bet is that scale, concentration and government support will keep it ahead of both.
Analysts are watching whether the commitment translates into pricing power or simply more supply. If AI demand keeps growing as expected, the new capacity will be absorbed quickly, and South Korea’s makers will emerge with a stronger grip on the market. If demand cools, the industry faces the kind of oversupply that has repeatedly punished memory companies. The country is betting that this cycle is different, that AI memory demand is durable enough to absorb the largest industrial investment in its history. The bet, at $550 billion, is the biggest South Korea has ever made.
